Re: [PATCH 14/24] arm64: dts: qcom: Update the pmh0110.dtsi for Glymur

[Krzysztof Kozlowski <krzk@kernel.org>]

On 08/10/2025 17:02, Krzysztof Kozlowski wrote:
> On 08/10/2025 16:31, Kamal Wadhwa wrote:
>> Hi Krzysztof, Dmitry, Konrad,
>>
>> On Thu, Sep 25, 2025 at 09:57:02PM +0300, Dmitry Baryshkov wrote:
>>> On Thu, Sep 25, 2025 at 10:34:52PM +0900, Krzysztof Kozlowski wrote:
>>>> On Thu, 25 Sept 2025 at 22:14, Dmitry Baryshkov
>>>> <dmitry.baryshkov@oss.qualcomm.com> wrote:
>>>>>
>>>>> On Thu, Sep 25, 2025 at 05:08:54PM +0900, Krzysztof Kozlowski wrote:
>>>>>> On Thu, 25 Sept 2025 at 15:34, Pankaj Patil
>>>>>> <pankaj.patil@oss.qualcomm.com> wrote:
>>>>>>>
>>>>>>> From: Kamal Wadhwa <kamal.wadhwa@oss.qualcomm.com>
>>>>>>>
>>>>>>> Add multiple instance of PMH0110 DT node, one for each assigned
>>>>>>> SID for this PMIC on the spmi_bus0 and spmi_bus1 on the Glymur
>>>>>>> CRD.
>>>>>>>
>>>>>>> Take care to avoid compilation issue with the existing nodes by
>>>>>>> gaurding each PMH0110 nodes with `#ifdef` for its corresponding
>>>>>>> SID macro. So that only the nodes which have the their SID macro
>>>>>>> defined are the only ones picked for compilation.
>>>>>>>
>>>>>>> Signed-off-by: Kamal Wadhwa <kamal.wadhwa@oss.qualcomm.com>
>>>>>>> Signed-off-by: Pankaj Patil <pankaj.patil@oss.qualcomm.com>
>>>>>>> ---
>>>>>>>  arch/arm64/boot/dts/qcom/pmh0110.dtsi | 66 ++++++++++++++++++++++++++++++++++-
>>>>>>>  1 file changed, 65 insertions(+), 1 deletion(-)
>>>>>>>
>>>>>>> diff --git a/arch/arm64/boot/dts/qcom/pmh0110.dtsi b/arch/arm64/boot/dts/qcom/pmh0110.dtsi
>>>>>>> index b99c33cba8860f1852231db33a127646c08c1e23..4a5c66e5c9fbc35cedb67601f4568844dc41fbea 100644
>>>>>>> --- a/arch/arm64/boot/dts/qcom/pmh0110.dtsi
>>>>>>> +++ b/arch/arm64/boot/dts/qcom/pmh0110.dtsi
>>>>>>> @@ -7,6 +7,8 @@
>>>>>>>  #include <dt-bindings/spmi/spmi.h>
>>>>>>>
>>>>>>>  &spmi_bus0 {
>>>>>>> +
>>>>>>> +#ifdef PMH0110_D_E0_SID
>>>>>>
>>>>>> NAK
>>>>>>
>>>>>> I already explained on IRC in great details why.
>>>>>
>>>>> A short summary or a link to a channel / date would be nice in order to
>>>>> include other people into the discussion.
>>>>>
>>>>
>>>> Of course but:
>>>> 1. You were there so maybe you remember the arguments, and:
>>>> 2. I'm offline, using phone, not having laptop, replying during my
>>>> personal time off just before merge window so any emergency time
>>>> should be spent on important matters instead these two huge patch
>>>> bombs adding such usage I already said: NO, don't do this.
>>>
>>>
>>> Well, If I'm asking, it means I don't rememebr the discussion. And I
>>> defeinitely didn't know that you are spending your personal vacation
>>> time in ML. And if the discussion was with some other people, then
>>> somebody else can drop the response to the question.
>>
>> Just wanted to give some background on this patch.
>> Even though PMH0104 and PMH0110 are common (b/w Kaanapali and Glymur),
>> they don't share the SIDs. So we tried to use status="disabled" to handle
>> this but we observed that because of the node name being common in the
>> two included files, it ends up overwriting the previous node with the
>> same name.
>>
>> eg-
>> #include "pmh0104.dtsi"  // assume contains pmic@4 { ...};
>> #include "pmh0110.dtsi"  // assume contains pmic@4 { status=disabled;};
>>
>> Here intention was to use the pmh0104 on sid-4, but it gets overwritten
>> with the pmh0110 on sid-4 ( with status disabled). This is why we ended
>> up using the `#ifdef`, ensuring that we can control the exact pmic that
>> gets picked by using the PMXXX_SID macro.
>>
>> side note, i did `grep` in the `/arch/arm64/boot/dts/` and i see a lot
>> of instances of `#if...` present in that.  Assuming the concern here is
>> about the use of `#ifdef`.
> 
> 
> #if are not desired in C code, so why would they be acceptable in DTS?
> It is not making the code easier to read at all.
> 
> On IRC in these older discussions I was very strongly against any DTSI
> which depends on some sort of outside values, except basic usage of
> defines. Original pmh0110.dtsi from kaanapali is fine:
> 	pmh0110_d_e0: pmic@PMH0110_D_E0_SID {
> 
> but doing ifdefs here that this define depends on something else makes
> code ungreppable (lookup unit address from sysfs and then git grep
> pmic@4) and difficult to follow.
> 
> My recommendation is either duplicate code or change DTSI files to not
> contain entire node, but its contents. At least these are
> recommendations I remember now.
> 
One more recommendation:
Different DTSI files per SoC, so pmh0110-kaanapali.dtsi and
pmh0110-glymur.dtsi.

We do not move the QUP engines into separate DTSI file to avoid
duplication. Each SoC has all of them duplicated. And then new SoC also
duplicates them.

We do not move SPI touchscreens or other I2C/SPI devices to separate
DTSI files. Each board duplicates such SPI device (and SPI can have
different <reg> - chip select! I2C as well, although a bit rarer).

Therefore I do not see any need for some #ifdefs in this code, just to
avoid duplication because that duplication is nothing odd.

Best regards,
Krzysztof